A pulmonary nodule is worth 8 × 8 × 8 words: Computed tomography-based 3D vision transformer predicts early-stage high-grade lung adenocarcinoma of micropapillary and/or solid subtypes

作者：Yin Zhou, Yiyang Wang, Cheng Li, Cheng Li, Shanshan Wang, Yuning Pan, Weiyu Shen, Chengbin Lin, Xuelian Ruan, Xianwang Ye, Zhenya Zhao, Zizhuo Wang, Jun Chen, Chenwei Li, Chenwei Li, Hui Liu, Wentao Fang · 发表于：Intelligent Medicine · 年份：2025 · DOI：10.1016/j.imed.2025.11.001 · 被引用次数：2 · 研究领域：Lung Cancer Diagnosis and Treatment、Radiomics and Machine Learning in Medical Imaging、AI in cancer detection

Early-stage high-grade lung invasive adenocarcinoma (IAC) has poor prognosis and is hard to identify using conventional radiological assessment. Current reliance on postoperative histology to identify high-grade subtypes delays risk-adapted surgical planning. 3D vision transformers (ViTs) may improve prediction by modeling long-range dependencies in CT scans. This study aims to develop and validate 3D-ViT and Swin Transformer for preoperative CT-based prediction of early-stage high-grade IAC subtypes (micropapillary/solid), benchmarking against ResNet. A multicenter cohort of 1028 patients with surgically confirmed early-stage lung adenocarcinoma was divided into training (n=806), validation (n=100), and external test (n=122) sets. 3D-ViT, SwinT and ResNet models were trained on CT to classify nodules harboring high-grade histologic patterns. A novel decision-aid tool for IAC surgery was provided. Performance was evaluated using AUC, accuracy, sensitivity, specificity, and precision. Attention mapping was performed to interpret 3D-ViT decision-making. The 3D-ViT model achieved AUC values of 0.856 (95% CI: 0.845, 0.877) (validation) and 0.806 (95% CI: 0.790, 0.816) (testing), compared to 0.854 (95% CI: 0.841, 0.872) (validation) and 0.760 (95% CI: 0.743, 0.776) (testing) for the ResNet baseline. 3D-ViT showed balanced accuracy, sensitivity, specificity and precision in validation set. In external testing, 3D-ViT significantly outperformed ResNet in all metrics with p < 0.01. T...